JPH03503461A - Error propagation image halftoning with time-varying phase shift - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] [Technical field] The present invention provides a multi-level display system having a relatively small number of brightness level display elements. More specifically, such related to adapting systems.

[Background technology] Flat panel display systems, such as binary liquid crystal displays (LCDs), use a relatively small number of Despite having elements that can only display temperature values, the CRT monitor type display system does not have many of the drawbacks found in In particular, binary LCDs use electron guns or vacuum tubes. unnecessary and can be made much thinner than a television monitor.

Since LCDs require low power and voltage, they generate less heat during operation, and therefore It is particularly suitable for high-density and portable applications. Its hardware is quite durable so you can display the same image for an extremely long time without worrying about damaging the elements. stomach.

However, CRT display systems have a very large number of brightness values between the minimum and maximum brightness values. The degree value can be easily displayed.

In fact, the input to a CRT display system is typically an analog luminance signal, The effect is that the noise associated with the signal limits finer differentiation of the input signal. quantized.

The benefits of this precisely quantized brightness capability are twofold. First of all, this display The position represents the gradation of the image relatively accurately. For example, a typical signal pair in a video signal The noise ratio determines whether an element on an analog television is displayed as one of 25 values. "restricted". However, since this is a relatively precisely quantized image, each element The display value will be an approximation of the displayed portion of the image. In other words, the display is Has good "gray toning" or "half toning". Second, Related to the first feature, this display has relatively good spatial resolution. each The elements track the brightness of the image well, so changes in brightness between adjacent pixels are also well represented. be revealed. Therefore, the spatial resolution of this display primarily depends on the physical distance between each element. limited by distance.

Binary LCDs and, more generally, LCDs with elements of a relatively small number of brightness levels. quantized display systems, some of which reflect the advantages of precisely quantized display systems. It has shortcomings.

For example, a binary LC'D element turns the corresponding part of the image either on or off. Therefore, it is expressed. If the image of the part to be expressed is gray, it cannot be expressed well. . In other words, LCDs inherently have poor halftone rendering capabilities. Also, ha Although it is related to its poor tone representation ability, LCD display devices have limited spatial resolution. is also relatively inferior. Each element is either on or off, so you can convert one image to another. To shade an image, turn some adjacent elements completely on or off. It must be approximated in some way. As a result, the shading becomes abrupt. too sparse or sparse spatially. In the latter case, the displayed image The image has a resolution worse than the spatial spacing of adjacent elements.

In the prior art, display devices with relatively coarsely quantized luminance level components, An input signal adapted to a system with display elements of densely quantized luminance levels. We have considered the issue of processing technology for display. As described below, this The basic problem addressed by these prior art techniques is The number of possible brightness values is greater than the number that the display element can take. death Therefore, the processing reduces the relatively precisely quantized luminance values of image elements to a smaller number. to one of the display element brightness values. To perform such a conversion Therefore, in most cases, the brightness values of display elements are different from the brightness values of image elements. become.

For example, when displaying a standard television image on a binary LCD screen, the brightness of each image element is The degree takes one of 256 values between the minimum and maximum brightness values, but the corresponding Display elements are either minimum or maximum values. Brightness between the 1st and 256th values Displaying the display element with the maximum value or minimum value for the image element with the This is a small error. The image element has the 128th value, i.e. the minimum brightness value and the maximum brightness value. When the display element has a brightness corresponding to an intermediate value, the display element is set to the minimum or maximum brightness. This is a fairly large error. The third value, i.e. the minimum value, Display elements at the minimum brightness for image elements whose brightness corresponds to a value close to This is a very small error.

For many years, attempts have been made to subjectively display gray images on display devices with binary elements. Processing techniques have been developed to address the above problems. these techniques All methods are based on the ability of the eye to coordinate the brightness of a number of closely spaced elements, either light or dark. It is based on the premise that, together, we recognize shades of gray.

U.S. Pat. No. 3,937,878 addresses black and white imaging in binary display systems. Remembers one method (dither) to be applied. Each image to be duplicated Divide into a matrix of pixels corresponding to the relevant cell of the display panel. Each display cell Specify a predetermined threshold value for . The threshold is in a certain pattern, typically 18 (4X4) Repeat for every square element, evenly distributed between the minimum and maximum image brightness values. There is. If the brightness of a given pixel is greater than the threshold specified for the corresponding display cell , the cell is turned on, and if it is small, it remains off.

In such a system, very dark areas of the image have a minimum of 16 square elements. Since the threshold is not exceeded, the displayed area is dark. Similarly, very bright areas are Since even the highest threshold is exceeded, all 16 square elements become bright. complete darkness In the brightness region just between the brightness and full brightness, 8 out of 16 thresholds Therefore, 8 elements remain bright and the remaining 8 elements remain dark. The eyes are in this small area. to recognize the brightness of gray.

Another method of halftoning by error propagation is faucet (Fave). ett) and Schraek, “Hardware using error correction” Halftoning Techniques lJsfng Error Correction) Js Proceedings of theSociety for Information Display (27(4) (1986), pp. 305-308.

In error propagation methods for binary display devices, the starting point is also a spatially quantized image.

Simple thresholding of the amount by which a display element exceeds or falls below its corresponding image brightness value geometrically, which will later be quantized into display elements, rather than simply discarding them as in the case of Add it to or subtract it from the values of nearby images.

Therefore, in the error propagation method, halftoning is performed by or by adjusting neighboring elements to compensate for the shortfall. .

The present invention takes error propagation one step further and applies the above concept to a mosaic color display. This applies to all locations. On mosaic color displays, immediately to the right of a particular element. element is not necessarily the same color as that particular element. Therefore, the error is However, it is not necessarily physically propagated to neighboring elements that have already been processed. Same color The nearest element for processing is the immediate neighbor of a given element, for example in the case of a mosaic of equal enveloping angle lines. It is an element below the element on the right. Therefore, it is suitable for such mosaic colors. In the error propagation method used, errors are diffused diagonally between elements. element is a video ・When processed horizontally according to the standard rask order of data, the hardware Implementing this method with Bell is more complex. In such cases, an element , and do not take it out when processing diagonally adjacent elements. It will be necessary to do so.

A major problem with conventional error propagation halftoning methods is the It is a matter of low ability. This is especially true when the display element has only binary capabilities. be. As the image darkens, the "r-on" elements become sparse, making each one very visible. Begins to stand. The method of error propagation along a single path involves linear or arrow Artifacts in the shape of the original can also be seen. An isolated "on" pixel in the dark region is in the bright region It degrades image quality much more than isolated "off" pixels.

In addition, as mentioned above, the troublesome problems found with mosaic color display devices Addressing this requires special adaptation of techniques developed for black and white images.

[Summary of the invention] The present invention provides an LCD type mosaic color display having a relatively small number of luminance value display elements. The device displays a color image with relatively precisely quantized luminance value element regions. Provides a processing method and system for The image display method according to the present invention is as follows. It includes the basic error propagation method for mosaic color display devices described above. As described above , the mosaic color display is patterned so that diagonal columns consist of single-color elements. , the error propagates diagonally in this method. However, this basic In the conventional method, the "error" propagated to the first element in the diagonal direction becomes 0, which is more uniform. Generally, it is assumed that it is a temporarily fixed constant that is independent of position. " In an additional feature of the invention called pixel interleaving, the "error" is Propagates to the first element on the diagonal, varying depending on the image or frame being processed do. The error propagated to the first element on the diagonal is also called the "preload value j.

More specifically, the "error" propagated to the first element of each diagonal is its "error" ' is incremented as each frame is processed until ' exceeds the maximum elemental intensity value.

Once the maximum element intensity value is reached, the error propagation is restarted by subtracting from the maximum value. Open.

For binary representation, the increment of the preload error value associated with each diagonal The "on" element moves spatially along the corner line. “On” along this diagonal The spatial movement of elements is "pixel interleaving." As a result, halftone The display is recognized. This is because if all preload values are equal, the binary representation As the number of images displayed increases, the temporally integrated sum of This is because it approaches a tone image. In this way, the eye can receive multiple input images for the same input image. If the processing is fast enough to integrate the displayed images, use the real images to create 2! Recognized display approaches the actual coordination tone of the input image.

The present invention also provides a subjectively high-quality representation of the image. -Fuft toning can be achieved. Furthermore, the present invention also provides a fast, sequential series of tables. It is possible to subjectively achieve high quality color halftoning using the displayed image.

Furthermore, according to the present invention, pixel incrementing provides an artifact in the error propagation method. You can lose things. For binary elements, the image brightness value plus the error value When exceeds the threshold, do not unconditionally give the element "On 1 display brightness (, the last The displayed image is displayed only if there are less than some number of "off" elements after the "on" element. Become “on.” However, even if the previous diagonal is is considered "on" when making decisions about elements of

[Brief explanation of the drawing] FIG. 1 is a flowchart of the processing method of the present invention.

FIG. 2 is a schematic diagram showing mapping of image elements in rows onto display elements in row a. It is.

Figure 3 shows a standard mosaic color table with equal angle of envelopment lines drawn through each display element. FIG.

Figure 4 shows a mosaic color drawing in which image elements are referenced corresponding to equal enclosing angle lines. FIG.

FIG. 5 is a horizontal schematic of the mapping of the mth row referenced diagonally in the display according to the invention; It is a diagram.

FIG. 6 is a diagram showing relative brightness values of image elements, display elements, and processing threshold levels. be.

FIG. 7 shows that an image element has one of 256 brightness values on a standard television display. The display element can display two brightness values on a standard binary LCD display. A diagram showing the relative brightness values similar to Figure 6 when one of the be.

FIG. 8, like FIG. 6, shows relative brightness values, and the image elements of the brightness factor (m, n). It is a figure which shows the mapping typically.

FIG. 9 shows the actual physical processing of the elements in the invention with reference to the diagonal matrix. FIG.

FIG. 9A shows that in an actual physical embodiment of the invention, a diagonal line is drawn from an element with value E. FIG. 2 is a schematic diagram showing how a direction propagates to neighboring elements;

FIG. 10 is a block diagram of an embodiment of the invention.

FIG. 10A is a block diagram of another embodiment of the invention.

FIG. 11 shows the results for the first element in the mth row for continuously processed images. FIG. 4 is a schematic diagram showing the variation of pre-selected error values;

[Preferred embodiments of the invention] The present invention applies error propagation to a mosaic color display to create a mosaic with equal enclosing angle lines. The error associated with the first element on each diagonal along which the error propagates on a screen display. It improves on the prior art by systematically varying the difference. diagonal The "error" propagated to the first element above, i.e. the same thing, but with the error pre-negative By systematically varying the load values, pixel interleaving can be In the case of radial elements, a spatial movement of the "on" element along the diagonal occurs. Konoto and the time integral of the displayed image for a series of consecutive images or frames is input. approaches the actual coordinated tone of the image. Additionally, isolated areas surrounded by dark areas By suppressing bright pixels, artifacts are removed. The present invention is based on equal enclosing angle lines. Although not limited to mosaics with a pattern of We will focus on the case of linear mosaics (.

The present invention provides (1) halftone for mosaic color display using error propagation; and (2) halftoning using pixel interleaving. It has the following aspects. In order to describe the invention, some preliminaries applied to error propagation It is necessary to explain the concept in some detail.

(a) Detailed explanation of preliminary concepts related to the error propagation method The method can be represented by the flow diagram in FIG. Figure 1 shows a portion of image 2. an image matrix 4 containing matrix elements 6 with discrete brightness values; It is divided into. The matrix indices Q and p are the image to image It shows the spatial relationship of the matrix. The value of each element of the image matrix is , for example, from a spatial barge run and an amplitude quantization bar scene of a CRT video signal. Can be assigned.

The brightness value of each input matrix element 6 is the brightness of the corresponding area of the image 2 to be displayed. is proportional to. The brightness values assigned to each image matrix element are discrete is finite and the number of representative brightness values is finite, and the number of possible display brightness values is There are quite a lot of them compared to the numbers. The number of discrete brightness values that each image element can take is defined as q.

The image matrix brightness values are then processed by the processor 8 into the display matrix 10 is converted to Each display matrix element 12 has a certain brightness. Each display matrix The image element 12 has a spatially corresponding image element 8. However, Therefore, the brightness value of each display element is determined by the brightness of the corresponding image map (IJ Fuchs element 6). A one-to-one mapping is performed from the value added with the above-mentioned error value.

The number of brightness values that each display element can handle is discrete and quantized, and is denoted by r.

As described above, the present invention is advantageous in that the number of brightness values that can be handled by a display element is Appropriate when there are fewer than the number of degree values. In other words, it is the same thing, but the value of r is q. It's small. Therefore, the processing results in relatively precisely quantized image elements. The luminance value of is converted into one of the luminance values of a smaller number of display elements. standard tv picture When displaying an image on a binary LCD screen, the image element intensity is one of 256 intensity values. the corresponding display element has one of two possible values (on or off) It is.

FIG. 2 schematically shows a horizontal row of image elements and display elements in FIG. It is. The arrows indicate that the mapping of the present invention is one-to-one between corresponding matrix elements. to show that Here, the processor 8 in Fig. 1 (not shown in Fig. 2) is Process in between. Such real-time processing by processor 8 prevents "error" propagation. Although not a necessary requirement for the section to apply to this presentation method, such a feature is preferred. This is a new example.

That is, the processor converts one image element brightness value to the next image element brightness value. mapped to display element brightness before arriving at the processor for power. The processor , if the mapping speed is slower than the input speed of the image element luminance values, then the processor Before processing, the backlogged value will have to be stored. child This is known in the art as ``frame buffering''. However, Therefore, in FIG. 2, the image brightness value of the first element in row Q is the image brightness value of the second element. is mapped onto its corresponding display element before reaching the processor. Row Q and The same is true for the remaining element sequences of all rows. therefore , the displayed image is the original input to the processor as a time sequence of image intensity values. corresponds to the image in time.

Therefore, by sequentially processing all the elements in row Q of FIG. It is displayed on the display device 10 as a series of elements 12 with varying degree values. Extending this By sequentially processing a series of images divided into image elements such as element 6 in Figure 1, can be managed, thereby displaying a series of displayed images 10 on the display device at approximately the same time. can be displayed. This display then changes at the same speed and the video cannot be displayed. Ru. The frequency of input images is defined as the frame rate. Each input image is the image element of the image The processing speed treatment above is just another sequence of brightness values corresponding to Since, the frame rate is the rate at which the entire image is This means that the speed is such that it can be physically processed and displayed.

A “frame” represents an “image” and is composed of image elements with brightness values. Please note. In addition, the "new" frame is constant as the display is replayed periodically. It can also consist of the same input signal as the previous frame, as when representing an image of be.

(b) Detailed explanation of the basic method of error propagation for mosaic color display FIG. 3 shows a pixel pattern of a mosaic color display with equal color lighting angle lines. R is A red pixel, G corresponds to a green pixel, and B corresponds to a blue pixel. These colors correspond to the three primary colors are doing. Due to the analog input signal, i.e. its inherent signal-to-noise ratio limitations An input signal with precise amplitude quantization is typically continuous in time and analog It consists of overlapping signals representing the brightness of each of the three primary colors representing a specific area of the image. Ru. Therefore, each spatial quantization of the input image, i.e. the temporal division of the input signal, and the assignment of representative brightness values to each time interval, corresponding spatially to the display elements. Three discrete inputs corresponding to the brightness levels of the eight primary colors of the analog image in the area There is an amplitude. These three input luminance values are of course subject to the inherent signal-to-noise described above. Due to the ratio limit, it is quantized discretely but precisely.

Therefore, mapping image brightness values to display brightness values is It's not easy. In the present invention, we consider two techniques for handling image brightness values.

The image brightness value is composed of three primary color brightness values that should be mapped to a display brightness value of one primary color. Ru. The first method is to calculate the actual image brightness values of two primary colors that do not correspond to the primary colors of the display element. It is something to be ignored. This method is advantageous when spatial resolution is important. No. Method 2 effectively divides temporally adjacent image brightness into groups of three, and Average the image brightness values for each primary color over an interval of Mapping the luminance value to a display element that spatially corresponds to one of the image luminance values with the primary color in question. It is something to be imaged.

This method is advantageous when accurate coordination tones are more important than spatial resolution.

The present invention integrates the handling of both input image signals. The invention is generally applicable Therefore, in the following discussion, we assume that the input signal has previously been processed according to one of the two methods mentioned above. and precisely quantized colors that correspond to the primaries of spatially corresponding display elements. Assume that the image brightness values determined are to be mapped to the display element.

With this in mind, we can apply the same mapping considerations discussed above with respect to Figures 1 and 2. , by a matrix of image elements with color intensity I(me n) A matrix of display elements with color intensities D(m,n) can be mapped to However, I (men) is obtained as a result at this time. Consider the image brightness of one primary color determined according to one of the two methods above. Must be.

In Figure 3, parallel monochromatic diagonals can be drawn from all elements of the mosaic. I understand. In Figure 3, the diagonal line starts from m=1 at the bottom left and goes up the left side of the mosaic. It is also numbered across the top edge. In the upper left corner, N　m　　　　　100 is on the left They are numbered diagonally through the top corner. Of course it's bigger than that. or a small height mosaic color display device, but choose 100 as an example. .

Figure 4 shows a matrix that refers to the monochromatic pixel diagonals of the mosaic color display in Figure 3. shows an image matrix with box elements. Mosaic cards improved with the present invention In the error propagation method for color display, the error is diffused diagonally between pixels, so the image Elements and Display Elements M) IJ Wax It is convenient to refer diagonally.

For this reason, the elements in the "matrix" in Figure 4 are similar to a standard matrix. , are not arranged in perpendicular rows and columns. Furthermore, each diagonal is not necessarily connected to any other diagonal. They do not have the same number of elements. As a result, Figure 4 and related drawings can be divided into m diagonals. 1 to nt for each diagonal element.

0. They are most appropriately numbered and referred to by n. However, see Figure 4. Since the elements shown have similarities with standard matrices, the explanation uses the matrix's Use terminology. Therefore, by numbering each diagonal, the first item of the matrix index, and each element from the "top" to the "bottom" of the diagonal is also numbered. Let those elements be the second index of the matrix.

Therefore, FIG. 4 shows a diagonal row numbered m, and the elements on the diagonal are Numbers such as n=1.2, . . . are attached to the bottom end. The image element is i (diagonal number , diagonal position) or 1 (me n). For example, the 98th The second diagonal element of is i (98, 2). Each diagonal also has the same number of elements. Please note that For example, row 1 has only one element, but row 5 has 5 It has two elements. Therefore, the maximum number of elements in a row, ntotml, is a function of m. Ru. As already discussed, each image element i(men) has a It has a brightness value I (me n).

Although not shown in Figure 4, the same table with matrix element d (m ln) The display matrix is referenced diagonally in exactly the same way as for the image elements in FIG. Ru. Furthermore, each display element d (me n) is associated with itself D (m +n).

Note that physical embodiments typically process each element in horizontal rows. Details are discussed below. In our method, where the mosaic has equal enclosing angle lines, the error is Since it propagates in the angular direction, the processor memorizes the error of a certain element and at the same time Process intermediate elements on horizontal rows, diagonally adjacent elements arrive for processing When doing so, it must be possible to correct the error.

In Fig. 5, the mth of the mosaic color image matrix and the display matrix The diagonal rows of are shown horizontally. The arrow between the image element and the display element It shows a one-to-one correspondence between trix elements and display matrix IJ Fuchs elements.

Image element 1 (me n) has an intensity value I (m, n). Display element d (m + n) is one of r possible values It has a brightness value D (mln). As mentioned above, the image brightness value I (me n ) is larger than the number that display brightness value D(m, n) can take. paraphrase If so, q>r. Define r display brightness values as A1, A2, ..., Ar. to justify Therefore, D(m,n) is equal to A1, A2,..., or Ar. It becomes.

To understand the invention, for each matrix element, the image element brightness I (my Let's consider a simple example of mapping n) to display element luminance D (m, rn) . Between each of the r display brightness values Al, A2, ..., Ar, a threshold T1, There are T2, . . . , T(r-1). Therefore, T1 is between A1 and A2, T2 is between A2 and A3,..., Tx is between Ax and Ax+1,..., T(r-1) is between Ar-1 and Ar0. Although the image brightness value I (me n) is larger than Tx When smaller than T(x+1), the display brightness value D(m, n) is A(x+1).

At the extreme value, if I (myn) < T 1, then D (me n) = A 1 , and if I (me n) ≧ T (r-1), then D (m, n) = A-r. be.

Figure 6 shows the q brightness values that %I (me n) can take and the values from A1 to Ar. r possible display values D (my n) over the range of and r-1 threshold levels T It shows the relative relationship between l, T2, -1T(r-1). Brightness value I (mt n) is normalized to range from 0 to 1. q possible values for each image element The image brightness value of is greater than the r display brightness values D(m, n) that each display element can take. . In other words, the image elements have relatively finely quantized luminance compared to the display elements. It has a degree value.

The basic feature is that the number r of display brightness values is smaller than the number q of image brightness values. There is an error associated with the mapping shown below. Figure 7 shows the case where the image element is a television screen. If the display element is a binary LCD display, The case where one of the two luminance elements is taken is shown, as in the case of In this case, q=256, n=2, A1=O1A2=1, and T1 becomes, for example, N　O-5,0. Choose sea urchin. All image brightness values between levels 1 and 256 have a brightness of 0 or 1 The dashed line indicates that the Therefore, by this mapping, the record Image brightness values between 129 and 255 will result in display elements being too bright and the level Image brightness values of 2 to 128 result in too dark results.

This error causes a relatively finely quantized luminance value to be compared to a relatively coarsely quantized luminance value. It is inherent in systems that map to values (not necessarily binary). No. FIG. 8 is an enlarged view of possible brightness values near the X-th image brightness value.

The Xth and (x+1)th image brightness values are between the threshold Ty and T(Y+1). Ru. Therefore, the image element brightness value I (m.

n) is at one of these brightness values, the display element brightness value D(m,n) is defined as According to the simple algorithm, it will have the mapped luminance value A(Y+1). Ru. On the other hand, the (x-1)th image brightness value is between Ty and T(y-1). death Therefore, according to this simple mapping algorithm, the image element of luminance level x-1 has display brightness (if!Ay).

As can be seen from Figure 8, if the image element is at the Xth brightness level, it will not be displayed. The brightness D (m, n) = A (Y + 1) is 21 from the image element brightness. increases by the amount shown. Similarly, the image element has the (x-1)th brightness level , the displayed brightness D (m + n) = Ay is 2 is reduced by an amount shown as 2.

In the present invention, this mapping error is compensated for using an error propagation method. What is error propagation? In general, the adjacency table due to over- or under-representation of the image element brightness I (m + n) This means that the value of the display element brightness is adjusted by the display element brightness D (msn). do. This over- or under-representation resulting from the mapping is propagated to neighboring image elements. It is an "error". The n-th element of the m-th row is generally written as (m+n), and is 1 ( Used with the same meaning as m+n).

The error propagated to the (m+n) elements of the matrix is E(men).

The error is applied to the adjacent element in the mth row, that is, from the (m, n-1)th element to (m+) In the error diffusion method propagated to the n)th element, I( m, n) is mapped to the display element brightness D (m + n), the image element brightness Overrepresentation or The amount of underrepresentation is subtracted from the image element intensity I(m, n), respectively, or added to it.

For example, the display element brightness corresponding to the second image element brightness value I (m, 1) of the m-th row Consider the mapping to the value D(m, 1).

As mentioned above, q image brightness values are larger than one display brightness value, so D ( m + 1) is usually larger than % I (m + 1) by a certain amount, or becomes smaller. If this error propagates to the second element of the m-th row, E (m, 2) = I(m, 1)-D(m, 1). If the display brightness value is larger than the image brightness value, Note that in this case, E(m+2) is negative. Also, small places is positive. Therefore, if the displayed brightness is too large, it will be propagated as a negative number. If the brightness is insufficient, it will be propagated as a positive number.

Next, the display element brightness D (m , 2), in this method, the image element brightness I (m, 2) and the propagation The sum of the errors E (m, 2) thus obtained is mapped to the display element brightness D (m, 2).

In other words, the excess display brightness of the first image element before being mapped to the second image element brightness value A surplus or deficit is subtracted from or added to the second image element luminance value. Ru.

The image element brightness value I(m) is mapped onto the display element brightness value.

2) and the propagated error E (m, 2) as the “adjusted image element brightness value. This is the adjusted brightness that is propagated to neighboring elements. The excess or deficiency of the displayed brightness relative to the value. In other words, E ( m.

3) = [:I (m, 2) + E (m, 2)] - D (m, 2). In addition, the The adjusted image element brightness values of the three elements I (m, 3) + E (m, 3) are shown in the corresponding table. It is mapped to the display element brightness value D (m, 3).

The remaining elements in the m-th row are similarly mapped. Therefore, for the nth element of the mth row, is the adjusted image element brightness value I(m.

n) 10E (m + n) is the corresponding display element brightness value D (m + n) ). Propagated error value E (m+n) = [1(m+n-1) +E (m, n-1) code D (m + n-1).

(C) Detailed explanation of pixel interleaving Second image element i (mth row in FIG. 5) Returning to m + 1), there is no error value E(m, 1), i.e. the same thing, but , E(m, 1)=O. The present invention further provides an error value E(m.

1) A system that can select any value between 0 and the maximum value that the image brightness value can take. provide a program. According to the invention, the adjusted brightness value of the first element 1 (m, 1) + E (m, 1) is mapped to the corresponding display element D(m, 1). Furthermore, the mth line The error propagated to the second element of is calculated from the adjusted brightness value of the first element to the corresponding display element. The value after subtracting the brightness value, that is, E (m, 2) = CI (m, 1), +E (m, 1)] - D (m, 1).

Formally, the processing for each element in each row of the pixel interleaving function of the present invention is It's the same. For the n-th element, I (my n) + E (me n) is D (m, n).

Also, E (m + n) = CI (m + n-1) + E (m, n-1) code D(m+n-1). However, E(m, 1) is the selected value. be.

In the "pixel interleaving" method, the error preload value E(mel) is Whether it's true or different from frame to frame, it changes every time a frame is processed. Ru. More specifically, the error preload value for each diagonal exceeds the maximum element brightness value. is incremented as each frame is processed until If exceeded, the maximum value is started anew by subtracting . Mosaic color with equienclosing angle lines – In a display device, for binary display, the increment of error associated with the first element of each diagonal is , their diagonally “on” elements move spatially.

If all preload values are equal, the time integral value of the display increases as the number of front child images increases. As the result approaches an accurate coordinated tone image. Therefore, processing is faster and the eye If multiple display images for the same input image can be integrated, recognition can be performed using existing images. The displayed representation approaches the actual coordinated tone of the input image.

Therefore, the present invention is summarized as a method of image display comprising the following steps: be able to.

(a) Providing an image including a plurality of image pixels i (me n). m is Contains integers from 1 to rn t o t a l, where n is from 1 to ntata ntotal is a function of m, and each image pixel t(me n ) has a brightness factor (m,n) equal to at least one of the q image brightness values. Ta where q is at least equal to 3 and each image pixel has one position.

(b) providing a display including a plurality of display pixels d(m+n); each The display pixel has a position corresponding to the position of image pixel i (m, n), and each display pixel d(m, n) is equal to one of r amplitude-ordered display brightness values A1, A2, ..., Ar It is possible to emit light with a new brightness D (m, n). However, r is an integer smaller than q. The number Ax is the Xth display brightness value.

(c) Defining r-1 threshold values Tl, T2, . . . , T(r-1).

However, TX is the Xth threshold.

(d) Defining the error function E (me n). However, E(m, n ) = I (me n-1) + E (m, n-1) - D (m.

H-1), and E(m, 1) is a function of time and only m for all m. Ru.

(e) For m: 1 to m t o t a l, n = 1 for each m Shin,. 1. , a screen that displays a display pixel d (me n) with a certain brightness is Tep.

(iLr (me n) + E (me n) ≦ When T1, D (mln) = A1 (ii) When I (me n) + E (me n) ≧ T1, D (m rn) = A r N or (iii) r>2 and T1≦I (m, n) + E (m, n) < T (r-1) When, D (m, n) = Axo, where X is a value between 2 and r-1, and the condition TX-1≦ I (my n) + E (m, n,) < TX is satisfied.

In the above discussion and the following discussion, we mainly focus on mosaic color with isochromatic diagonals. Although the emphasis is on display devices, the invention is not limited thereto. . The present invention provides displays with equal color rows or columns, as well as hexagonal coordinate patterns, etc. Suitable for displaying other heel variations. In any case, each element is expressed in matrix notation It is expressed as (men). where m is based on a specific mosaic pattern Referring to the grouping of elements to be processed, n is the processing of elements in the mth group. Refer to the process order.

Furthermore, the sequence of input signals does not necessarily correspond to the mapping sequence of elements. There is no need to By the same token, the mapping need not be done in real time. parable For example, if the input signal for a certain frame is stored in a memory memory, They can be accessed in a particular order of processing.

This error diffusion method for equal-enclosing angle line representations isolates errors from necessarily one display element to another. There is no need for direct propagation to adjacent elements within the corner line. A simple extension of the method described above is , divides the error from one display element and spreads it diagonally to multiple neighboring elements. can be done. Therefore, when processing the n-th element of the m-th row, the error value E ( my n) is the error from some elements in front of the mth row multiplied by a certain percentage. equals the sum of things. For example, the value of E (m, n) is From 1/2 of the sum of the adjusted brightness of the elements, the brightness value of the display element corresponding to them is calculated. It is equal to the value obtained by subtracting 1/2 of the sum. The same thing is N　E　(me　n ) is equal to E(mln-2) code D(mtn-2).

Another extension method uses the error E (m y n) not necessarily only on the mth diagonal. , and can be spread to multiple neighboring elements. Therefore, the mth diagonal When processing the nth element of the line, the error value E (m, n) is the mth, (me3) )th, (m+6)th, 91. The error from multiple previous neighbors on the diagonal of It is equal to the sum of times multiplied by a certain percentage. This is because their diagonals are simple This is because in a display with a color diagonal line, it has the same color as the (me n)th element.

For example, E (me n) + E (m + 3. n + 2) code D (m + 3. n +2) is equal to ko. However, element (m+3.n+2) is the same as element (m, n) is the next closest already processed element of color.

In the above “multi-element” and “multi-branching” methods, E (mt n) = ΣΣK (me m’nl j)X mouth I (m’+ j)+E (m’+ 　J)　　D　(m", j).

however, (i) m” is m t o t to be processed, 1 for i element groups Vary between references. j is between references to each element in a group of m" elements. change.

(u) K (me m’ + nl j) is i (m’, j)? is a propagation coefficient for error propagation from i (mln).

(iii) E(my1) is a function of time and m.

K(m, m"+n+ j) is Note that is O.

Furthermore, the above formulation is not limited to mosaics with isochromatic diagonals; Note that it is equally applicable to mosaics of other patterns. each element must of course be expressed in matrix notation (m, n). however, m refers to a group of elements to be processed based on a particular mosaic pattern; n refers to the processing order of each element in the mth group.

Figure 9 shows the diagonally referenced rows of the matrix model in a typical physical This is related to the horizontal processing of the embodiment. The diagonal rod m in FIG. and one of the monochromatic diagonals of the mosaic color display along which the error propagates. handle. The first image element of the mth diagonal rod is physically processed in the horizontal row. is the first element. E(m, 1) is preselected as described above. Then, I (ml 1) + E (m, 1) is the corresponding table in physical processing. This is processed by the display element D(m, 1) (not shown in FIG. 9). Then physical processing Therefore, the image element at the horizontally adjacent position is the (m+1.1)th element. mapped to This is because, as mentioned above, images and display electronic signals are typically This is because it corresponds to quasi-rusk display. Physical processing is the last image on that horizontal row It continues until an element is processed, and once it is finished, the next horizontal row of processing processes the image element. It starts from (m-1,1). Only after that, the image element (ml 2) The second element of the N or mth diagonal rod is physically processed.

As is clear from the above description and FIG. 9, some elements in a horizontal row In the physical embodiment, processing is performed between adjacent elements on a diagonal rod. did Therefore, this physical example has the value [I (m, 1) + E (m, 1) code D (m tl) = E (m, 2) and access its value accurately to create multiple intermediate images. After processing the image element % I (mt 2) + E (m.

2) It is necessary to have a means for mapping. One way to do this memorization is by horizontal It uses a row buffer with storage capacity equal to the number of elements in the row. Mistake Differences must also be propagated to and from each intermediate element on that horizontal row; Since the intermediate elements of these are also on other diagonal rods, the row buffer of horizontal size is suitable for this function.

The row buffer is a FIFO soft register of size equal to the number of elements in the horizontal row. It does this in a manner functionally similar to . Referring to Figure 9A, in the horizontal direction The resulting processed element 21 before processing of adjacent elements 22 begins. The value E (m, 2) is loaded into the buffer 20. Each intermediate rask order element 23~ 28 is processed, the value of E(ml2) is directed to the output 26 of the buffer 20. At the same time, data corresponding to the errors of intermediate elements 23 to 28 are inputted. , drawn from the buffer 20. The second element 29 of the m-th diagonal rod is physically When processed, the value E(m,2) is at the output 30 of the buffer 20 and is You can access it for any reason. After processing, the resulting The error E (m, 3) (not shown) that is propagated to the element is input to the buffer 20. Ru. Figure 9A shows a display with seven intermediate elements, but with more or less The same is true for display devices with large horizontal elements.

Referring to FIG. 10, one embodiment of the invention is shown. This example is When propagating errors and processing intermediate elements according to the method of the present invention, errors can be stored in memory. have the means to do so. The order of input and output signals is to the left of the standard rask scan line. Since it corresponds to the order from to the right and from top to bottom, the element N CQ, p) is It is easiest to think of them as elements that correspond to horizontal and vertical columns of image elements. . To put it simply, the image element brightness value I (Q, p) of the image element CQ, p) is the input It is input by force means 42. If Ω=1 or p=1, this reference for each element In the scheme, the element is at the upper end of the diagonal rod and according to the invention the error value E (Q , p) must be input. Therefore, access to preload buffer 52 is access and retrieve E(Ω, p). If the element under consideration is not at the top of the diagonal , that is, if standing≠1 and p≠1, the appropriate error value E　CQ, p) is stored in the error memory. It is removed from the means 48. E l.

No matter how p) is obtained, it is added to ICU, p) in the processing means.

The threshold determining means 46 is accessed by the sum I (Q, p) + E (Q, p). Ru. The threshold value determination means 46 determines that the sum I (Q, p) + E (Q, p) is the above r-1 number. Determine which of the thresholds it is between. The processing means 50 uses an appropriate threshold value to The display brightness value D (Q) is output by the output means 44 based on the threshold value.

p) is output. To reference elements horizontally and vertically, i CQ, p) The diagonal element for is i (fL+1.p+1). Therefore, the processing The stage 50 has the value E (ml1.1) + 1) = [:In, p) 10E (L) p) Code D (Q, p) is also stored in the error storage means 48.

The error storage means may be a row buffer of size equal to the number of vertical column elements of the image element. .

The above device can generally be applied to devices that propagate errors diagonally. , and is not limited to the specific mappings mentioned above. This device is ICU, p ) and E (Q, p) can be used for any determination of <D CQ, p), and always However, it is not necessary to use the threshold processing method of the present invention. Furthermore, ICU.

El+1.1)+1) Throat determined from p), E(look, p) and D(breath, p) It can be used for any value.

More specifically, this particular embodiment includes the following means.

(a) Multiple image pixels iCQ, each corresponding to a certain position on the image, p) A plurality of corresponding luminance hooded signals ICU, p) from the standard left to right, top to Input means that Uke takes in rask order to the bottom end. However, Q is from 1 to Q. talma Contains an integer of , and corresponds to one horizontal line Q tota from the top end to the bottom end of the rask scan. and p is 1 to p. Contains integers up to t8□, pt from left to right of rask scan c+ta corresponds to one vertical column. Each luminance coded signal I (Q, p) has q pieces , and q is at least equal to 3.

(b) A plurality of luminance coded signals D(12 , p) sequentially. Each display image d CQ + p) is the number of image pixels. Corresponding to position i CQ, p), each display brightness coded signal D(, Q, p) is. r It corresponds to one of the amplitude order display brightness values A1, A2, . . . NAr. however , r are integers smaller than q, A, x is the Xth display brightness value.

(c) Store error value E (1,, p) corresponding to the luminance hooded signal at the input error storage means.

(d) Pre-selected error value E corresponding to the number of luminance coded signals in the input l, p).

(e) By the following operations 1. Then, the luminance coded signal I (Ω7.p) at the input is Processing means for mapping to a luminance coded signal D(U,p) at the output.

(1) Take out the value I (fl, p) from the input means.

(2) If fi = 1 or p = 1, put the value E(Q, p) in the preload buffer get from

(3) If Q≠1 and p≠1, obtain the value E(Ω, p) from the error storage means.

(4) Value I CQ, p) and value E1. Determine the value D(fL, p) based on Set.

(5) Send the value D(ill, p) to the output means.

(6) Value based on I (stand, p), D CQ, p) and E ([,, p) Calculate (Q+1.p+1).

(7) Store E (rise+1.ri+1) in the error storage means.

Further, the output means is configured such that each element has a luminance value Al, A2, . . . , Ar of 2. Kotono It will be connected to a display device that displays color mosaic patterns.

Another embodiment of this device is to convert the error to that of its diagonal or other diagonal. It propagates to two or more adjacent elements. This device includes the following means:

(a) 2) Multiple image pixels 1 (Lp), each corresponding to a certain position on the image Corresponding multiple luminance coded signals (Ω, p) from standard left to right, top to bottom Uke takes input means in rask order to the end. However, Q is from 1 to Q. up to tal From the top to the bottom of the rask scan, including the integer Q! (Supports 1181 horizontal lines) and p contains an integer from 1 to p7.1, p from left to right of the raster. t corresponds to al vertical columns. Each luminance coded signal generator (breath, p) has q image luminances. corresponding to at least one of the degree values, q is at least equal to 3 and 2.

(b) Multiple luminance coded signals D ( Output means that sequentially outputs 0, T)). However, each display pixel d (fl, p ) corresponds to the position of image pixel i (11, p), and each display brightness coded signal D ( Q, p) corresponds to one of the amplitude order display brightness values AI, A2, ..., Ar. Ru. However, r is an integer smaller than q, and AX is the Xth display brightness value.

(C) Luminance-coded signal processing (Ω, p) at input from previous element (Q’, p’) ) corresponding to the partial error value PE (LQ').

partial error storage means for storing p, p').

(d) Pre-selected error values E corresponding to multiple luminance coded signals in human input Buffer preload means to maintain (11, p).

(e) Output the luminance coded signal generator CQ, p) at the input by the following operations: processing means for mapping to a luminance signal at.

(j) Take out the value I (CL, p) from the input means.

(2) When Il=1 or p=1, the value E ( Ill, 1)) is obtained.

(3) When Q≠1 and p≠1, the value PE (11, Ω' , p, p"), and sum the values of PE (Ω + Q' + p + p") to get E ( fl, p) is obtained.

(4) Based on the values of I (Ω, p) and E (Ω, p), (iiD(L) Determine p).

(5) Send the value D([, p) to the output means.

(6) Calculate the partial error value PE(a, 91 blp). However, a and b) are This is an element to which the error is propagated from (Q, p).

(7) For all (a, b), the partial error value PE(a, Q, + b +p) is stored in the partial error storage means.

In this embodiment, the partial error storage means may be a plurality of row buffers.

More generally, the device is capable of processing input signals that are not necessarily in standard rask order. The receivers take the input signals in a different order and map their input signals according to a different order. In that case , as shown in FIG. 10A, between the input means 42 and the processing means 50. , requires a storage matrix 54. Therefore, such a device must: It will include the means.

(a) Receive multiple luminance coded signal lights (Lp) corresponding to a certain position on the image. is the input means to take. However, Q includes an integer from 1 to 9.1°tl and is not processed. Q. Referring to t1 element groups, p is 1 to p. , up to 11 contains a number, refers to the elements in the Q-th group according to the processing order, and l) tゎ, 1 is , m. Each luminance coded signal In, p) is a small number of q image luminance values. corresponds to at least one, and q is at least equal to three.

(b) Store the input image brightness value of the complete frame t5, and access the memory t5. Trix memory device.

(c) Multiple luminance coded signals corresponding to the position of input image luminance value I ([,T)) An output means that sequentially outputs the numbers. However, each display brightness coded signal Dl,1)) is Corresponds to one of I amplitude order display luminances (i! A1, A2, . . . , A r).

r is an integer smaller than q, and Ax is the Xth display brightness value.

(d) corresponds to the processed luminance coded signal (Ω, p) from element (Ω’, p’) Partial error value PE CQ, 0. partial error storage means for storing t, p, p') .

(e) Pre-selected error values E corresponding to multiple luminance coded signals to be processed l, p);

(f) By the following operation, when outputting the luminance coded signal I (fl, p) Processing means for mapping to a luminance signal.

(1) Take out the value I(Q, p) from the matrix storage means.

(2) If Q=1 or p=1, the value E1. p) is obtained.

(3) If Q≠1 or p≠1, for all Q゛ and p“, the value PE i, n', p, p') are obtained from the partial error storage means, and PE (Q, ( 1'', p, p') to obtain E(Q, I)).

(4) Based on the values (see, p) and ECU, p), the value D l.

Determine p).

(5) Send the value D(cl, p) to the output means.

(6) Calculate the partial error value PE (a, Q'l bl 1)). However, N (a-b) is an element to which error is propagated from (IQ, ])).

(7) For all (a, b), the partial error value PE(a, + Ω, b, 1)) is stored in the partial error storage means.

The above device can also be used for patterns other than mosaic color patterns with isochromatic diagonals. Applicable. The elements are referred to by the matrix notation CQ, l)). However, Q defines a group of elements to be processed based on a specific mosaic pattern on the display device. where p refers to the processing order of elements within the group.

In the following discussion, as before, the reference is the diagonal of a mosaic with isochromatic diagonals. shall correspond to the line. In other words, m is m of the mosaic color display device. t o + a One of one monochrome diagonal line, n is the nth point from the top of the diagonal It is basic. Again, as an example, we will focus on a mosaic with °C isochromatic diagonals. However, the present invention applies generally to mosaic color patterns.

E(m, 1) for each diagonal rod corresponds to an image element that lies on the physical boundary of the image do. Since element (m, L) is the starting point of the diagonal rod, the propagation of error from the previous element is do not have. Therefore, as mentioned above, E (m, + 1) is for each m diagonal Can be selected for rods. Its value is, for all m diagonal rods, They may be the same or different. Moreover, it may change for each image.

The error value E( As a result of changing m, 1), any of N D (m + n) for all m or everything changes. Referring to Figure 1, it says, ``Row j is drawn horizontally. However, as in the previous drawings, the diagonal rod m of the mosaic color display device is It may correspond to any mosaic pattern, or it may correspond to all mosaic patterns in general.

These two frames are the same, or at least the mth of two consecutive frames. Assume the rows are the same. Also, in this explanation, N E (m + 1) is The two values between the minimum image brightness value and the maximum image brightness value for consecutive frames of Assume that they are taken alternately. Since the values of E(m,,1) are different, continuous diagonals The errors propagated to the elements are generally different. That is, the (m+n)th point Error related to the element E (m + p,) = I (m, n-1) + E (m, The values of n-1)-D(mtn-1) and I(m+n) are between image elements. This is because they are assumed to be the same. Furthermore, display element value D(m!n) may be different for corresponding images on two same frames. That's what I mean is N D (m + n) is the adjusted brightness value of the n-th element I (m, n) + mapped from E (m, n), and whether E (m, n) is different in two identical frames. It is et al.

Therefore, when E(rr+, 1) changes, for all m, D(m, It is clear that any or all of n) may vary.

The two brightness values for the same consecutive frames are expressed as the n-th display element brightness value D (m, n ) are taken alternately, and if the alternation is fast enough, the average value of the two luminances will be recognized. Ru. If the threshold level is approximately halfway between the two possible display brightness values, the threshold Image elements with nearby brightness values I (m + n) are It is not well expressed by values either. If two adjacent display brightness values are taken alternately ( In this method, the threshold level or image element brightness value is approximately equal to the new brightness is recognized. Furthermore, the luminance I(m + For an image element with n), successive frames are The display brightness is less likely to change between frames. This also gives good results . This is because if I(mtn) is not near the threshold, the possible brightness display level This is because it is relatively close to Ax and can be well expressed by mapping to AX.

Expanding the above explanation, the error E (m yl) value of the first element of the mth row is Vary between three or more values within the range of possible image brightness values for the frame be able to. For a series of identical frames, for each image (m, ri can be As the number of values increases, the average brightness D (mrn) of each display element increases with the corresponding approaches the value of the image element brightness I (m, n).

In other words, for each frame, each m is randomly chosen and 1.000 identical Process and display one frame with E (mt fl), and display these 1,000 frames. If the display is processed in such a short time that changes in the display cannot be visually distinguished, The display is identical to its frame in all respects, especially in terms of gray toning. It is recognized that

Although the above is the basis of the present invention, the processing speed is so high that it cannot be detected visually. is not in a state to process a large number of frames at this speed. Same at current processing speed If E(m+1) varies between 1,000 values for an image, it is visually rather than recognizing the average luminance of the resulting individual display elements on row m. changes in brightness are recognized. The specific example below addresses these two conflicting requirements. This is an attempt to reach a compromise. That is, for consecutive frames, E(m, 1) is changed to enable more accurate brightness display, and at the same time, the displayed E(m , 1) are kept unchanged.

In one embodiment of the invention, each m is initially uncorrelated with the mth row difference value.

Each E(m,1) has an initial value between O and the maximum image brightness value. E( m, 1) is incremented for each successive frame. E(m, 1) is the maximum image brightness value If the maximum value is exceeded, the maximum value is subtracted and processing continues.

In a more specific embodiment of the invention, the display element has two possible brightness values (in accordance with the invention). If expressed by the above general formula, only one of r=2) can be taken. A1 and A 2 are normalized to 0 and 1, respectively, and T1 is selected to be 1/2. Image brightness value corresponds to the value of the TV input, and there are 258 values of normalized brightness between O and 1. You can take one of the following. For each image, E(m, 1) for each m is Initially, arbitrarily select 0 or 0.5. For consecutive images, E( m, 1) takes O and 0.5 alternately. This is a good quality half-tone This is the preferred embodiment because it provides a

The advantages of the pixel interleaving function of the present invention are processed according to the above embodiments. This can be demonstrated by considering a uniformly dark (not black) region of an analog image. image element is smaller than T1 (i.e. 1/2), most display elements are turned off. It is f. If m-th line preload value E (m, Q) = O, then 1 (m, n) 10E (my n) becomes larger than T1, and (m.

Errors gradually accumulate between the elements in row m until the n)th element is "on".

However, while D (my n) = 1, I (my n) + E (my n) is only slightly larger than T1 or 1/2. Therefore, (m, n The error propagated to the +1)th element is approximately -1/2. Therefore, the following The error must accumulate to more than 1/2 to produce an "on" element. , there are a relatively large number of "off" pixels adjacent to (m, n) in row m. the result, On row m, "on" elements occur sparsely and evenly spaced.

In the next frame, E(m, u) is modulated so that E(m, u) = 1/2. , the "on" elements of that display are elements that are equidistant between the "on" elements of the previous display. Become basic. This means that all elements of the previous frame with an accumulated error of 0 now have an error 1/2, resulting in an "on" pixel, and the previous "on" element with an accumulated error of 1/2 is , with an error O, resulting in an "off" pixel.

Therefore, E (my a) alternates between 0 and 172 for consecutive identical images. In the case of Moving and visualizing the sparse time on a black background as in the case of constant preload A stationary "on" element will not be recognized, and the spatial and temporal average of the "on" element will be recognized. be done.

Another method is to choose the same parameters as in the example just described. However, each m E(m, fl) for is initially randomly selected between 0 and 1. continuous In the frame, E (m, 41) for each m is the initial value and (i) the initial value is If it is 172 or more, the value is 1/2 smaller than the initial value, (ii) the initial value is 1/2 uncleared. In the case of , one of the values 1/2 larger than the initial value is alternately taken.

Another feature of the invention is a method of suppressing artifacts that are unique to the invention. artificial Object removal is known in the art from other methods of gray toning. However, it does not work well with pixel interleaving. Artifacts may appear in dark areas of the displayed image. contains sparse "r-on" pixels that appear in the region. Each of these “on” pixels are very noticeable and reduce the quality of the displayed image. Such artifacts are books This is a natural result of the invention's error propagation method. This happens much more than the first M value T1. Even in image regions with low luminance, when the error is propagated to successive elements, This is because the error value of adjacent elements increases.

Finally, the adjusted brightness value of the n-th element E (m, n) + I (my n) exceeds T1, and D (m, 2) = A2. That is, within a uniformly dark image region , resulting in a glowing display pixel.

In the present invention, when the last display element in a row is displayed with a brightness greater than A1 This artifact can be removed by maintaining a record of the counter variable C. leave (1) The display element under consideration is , mapped with luminance A2, and (2) the last element with luminance value greater than Al. The display element with the brightness value A1 has been processed by a pre-selected number N or more since the start. If the record indicates, the display element is mapped with the brightness value Al rather than A2. both If the condition is not met, the display element under consideration is It is mapped with luminance A2, which is the result of the brightness.

Of course, whenever a display element is displayed, it is also displayed with a brightness greater than A1. If so, record C is reset using the nominal algorithm. similar When the display element under consideration starts a new line, record C is reset. . When removing artifacts, all treatments must be carried out in the manner detailed above. Let me emphasize that. The artifact removal function displays display elements in A1 or A2. A final decision on whether to display will be added.

Other minor aspects of the invention are self-evident. For example, in one embodiment, the display The luminance values are equally spaced, and the thresholds are equidistantly spaced between adjacent luminance values. It will be done.

Furthermore, this error propagation method with pixel interleaving can also be applied to monochrome display devices. can. In that case, there is no need to propagate the error diagonally. That is, horizontal This is because the elements adjacent in the direction are white and black. Errors can propagate horizontally. and elements can be referred to such that m corresponds to horizontal rows and n corresponds to vertical columns. death Therefore, for every i (my 1), i.e., for which the error is The selected elements will constitute the first vertical column of the image.

When quantizing the image display output, for example in some liquid crystal display devices, the output can be quantized both in terms of amplitude and spatial position. If the output brightness is If it is limited to another level, e.g. a pixel is either on or off and in between LCD displays that do not have brightness levels (also called pie-level displays) The indication is quantized in terms of amplitude. Each region of the input image is a discrete pixel in the display. When represented, the representation is quantized with respect to spatial location (i.e. the representation is pixel ). As discussed above, these quantized display media Continuous tone images are approximated using various halftoning methods.

As previously mentioned, temporal techniques can be used to create halftone images, e.g. pixel ink. Improved amplitude and spatial resolution of halftone images created by reaping It is possible to

Error correction using temporal integration is called "temporal halftoning." In some cases. Well, let me explain in a clever way 1. The resulting technique is mainly based on linear error propagation. It is aimed at methods.

The principle of the temporal halftoning technique of the present invention is that the temporal resolution of the human visual system is Based on the limited availability of display systems and most quantized display systems.

According to the invention, in successive time periods, the display device displays a variety of continuous tone images. , presents an equally valid amplitude quantized representation. These expressions are 1. high space They differ from each other only in the details of their frequency, and they cooperate and compensate for each other. Continuous tone image input A finite number of image representations can be iteratively cycled until the input changes. When this occurs, processing is restarted for a new input image.

According to a preferred embodiment of the invention: D (me nt jx) = G [(I (It t)) * All b≦n)] D (me nl tz) = G [: (I (me b) + P (my b).

All b:an) In the above equation, G is a single branch error diffusion function.

A is the minimum brightness step for a pixel in D, and if b=1 then P(m , b) = 0, 5" If Ab>i, then P (m, b) = O. Ru.

D (me nl t,) is the display level at a specific time t1, m and n are the column and row of a particular pixel, respectively.

The temporal halftoning technique can be compared to the more general static halftoning method. For example, to accommodate the aforementioned technique where the error propagates to two or more adjacent elements, Decide the display level D (mrn) for the image in the second and subsequent time periods. It would be convenient to define a general function for determining The above technique is the basis It would be particularly advantageous if it were independent of some static halftoning scheme. like that The function can be written using the following formula. In these formulas, the analog image (I (at b), all a, and all b) are one set of Z. Digital expression (D (m e rl + t t), D (me 1'l + tz)... This is well represented by the iterative sequential representation of D

These expressions are calculated as follows.

D (my rl + tz) = F Cm* nr (Z’ I (a , b) - '5i' D (at b + ty) + all a, all b))y+1 In the above equation, F is a single static representation (rendering) of an analog image S (m + n), i.e. S (me n) = F (mt nl (Work (at b) + all a.

all b)) is any halftoning function suitable for creating .

Of particular importance is the case Zo = 2 (in this case, temporal artifacts can be avoided). (This is particularly convenient since

D (me nl t,) = F [me nl (I (at b) + All a, all b) D (m, n, tz) = F [me nl (2”l (at b) - D( a, b + tt) + all a, all b)] Function F is a set of specific pixels In some cases (e.g. in the technique mentioned above), in other cases it depends only on a few pixels. It may also represent a calculated value.

To apply the above formula to one-dimensional error diffusion, simply use it and let F=G: good. Here, G is a one-dimensional error diffusion function. In such a case, the expression (with the meaning ) can be simplified as follows.

D (me n, tt) = G (I (me b) + all b≦n) D ( m, n, tz) = G (2xI (mt b) - D (m, b.

tl), all b:a 11) The halftoning function used to calculate the frame is F(x).

However, X is the level of the original image (I(a.

b), all a, and all b). (I(a.

b), all a, all b) are pixels of any or all pixels of the original image. Represents the image level. In the general case (I(a.

b), all a, all b) also have the special case I (me Contains n). The halftoning function F(x) is When using a one-to-one dependency between the display at specific times t1 and tz A more specific formula for calculating the level is as follows.

D (me nt tt) = F [I (me nl tt) D (m , nl tz) = F[:2''I (m, n, tt) - D (m.

n,t,)ko A good halftoning function F(x) is D(m, nl tt) and D The difference between (me nl tz> is defined above for any m and any n) guarantees that it is not greater than A, as shown in the table below. If that does not hold, then the given D (me nl, tt) and D7 (where the difference is larger than A for m and n) m + n + t2), their average value −0.5” (D (me To replace it with n, t1) + D (me rl + t2)) As a result, image expression can be improved.

Continuous order a (i.e. precisely quantized) image frames can be created using conventional electronic circuits. We run the above algorithm in real time using the normal rask order output of the system buffer. It is possible to build a machine that does this. Considering efficiency, D (mu nI The calculation of D(mu nI tz-1) is performed in the calculation of Using memory buffers to avoid re-running these calculations That is Yoizuki.

In one special case using spatial error diffusion and Zo''2, a very simple alternative is Alternative architectures can be used. Result D (m u n, t+) is retained. This buffer is only log2(g ) bit. where g is the number of display gray levels. D(mu n, Reading this buffer during the generation of A selection can be made from one of g lookup tables. D (m, n I (mu t+) and D (mu nI tz) n) and 2*I (mu n) are generated by the quantum adder of I (mu n) This is done by simple multiplexing by shifting the provision to (D( While generating m, , nl t+), D( use the lookup table used to generate my nI tz) ). By using sufficient lookup tables, e.g. Arbitrary quantum Note that it is possible to implement a

Figure 3 Note table flat 3-503461 C15) Ward Summary Figure 5 international search report Country@Investigation Report

Claims (7)

[Claims] 1. (a) a is an integer from 1 to atotml, and b is from 1 to btotml and btotml is a function of m, each image element i(a, b) has an intensity I(a,b) equal to at least one of the q image intensity values, and q is at least equal to 3, and each image element has one position. providing an image containing the prime i(a,b), (b) r is an integer smaller than q; Assuming that Ax is the Xth display brightness value, each display element d(m, n) is the image element i (m, n), and each display element d(m, n) has r amplitudes. Emit light with a brightness D (m, n) equal to one of the sequential display brightness values A1, A2, ...Ar. a step of providing a display device including a plurality of display elements d(m,n), which can floor, If x is a function of {I(a,b) all a, all b)}, m, and n a step of defining a halftoning function F(x); D (m, n, t1) = F [m, n, {I (a, b) all a, all b} Therefore, the step of calculating D(m, n, t1), D (m, n, t2) = F [m, n, {2 = I (a, b) - D (a, b, t1) all a, all b}] to calculate the subsequent value of D at time t2. Including, how to display images. 2. A fixed analog image is created by a series of non-identical digital representations that are continuous in time. How to display images. 3. 3. The method of claim 2, wherein a particular display array element of the representation is a predetermined half. An image with different display brightness values for each digital representation according to a toning function. Image display method. 4. 4. The method of claim 3, wherein a particular one of said display elements has a particular halftone. has a display luminance D(m,n) for a lighting function F(X), The display element is configured such that D(m, n, t1)=F[m, n, { I(m,n)}], The display element is configured such that D(m, n, t2)=F[m, n, {2#I(m,n)-D(m,n,t1)}] It has a display brightness calculated by If the image is for image element i(a,b), I(a,b) all a, all b of An image display method with a brightness of 5. 4. The method of claim 3, wherein one of said specific display elements has a display luminance D(m,n) for a lighting function F(x), The display element is configured such that D(m, n, t1)=F[I(m, n)], and has a display brightness calculated by Said display element at a subsequent second time t2 becomes D(m, n, t2)=F[2 #I(m,n,t1)-D(m,n,t1)], and the front An image display method in which the image has a luminance value of I(m,n). 6. 4. The method of claim 3, wherein a particular one of said display elements has a particular halftone. has a display luminance D(m,n) for a lighting function F(x), The display element is configured such that D(m, n, t1)=G[{I(m , b), all b≦n}] It has a display brightness calculated by The display element is configured such that D(m, n, t2)=G[{I(m ,b)+P(m,b), all b≦n}] It has a display brightness calculated by G is a single-branch error diffusion method, If b=1 then P(m,b)=0.5=If Ab>1 then P(m,b)=0 and A is the minimum brightness step for a pixel, and the image is I (a, b) all a, all b An image display method with a value of . 7. 4. The method of claim 3, wherein one of said specific display elements is displayed at any time t2. The display brightness D(m,n) for a specific halftoning function F(x) and and D (m, n, tx) = F (m, n, {Z*I (a, b) - ▲ mathematical formula, chemical formula, There is a table etc. ▼D (a, b, ty), all a, all b}), Z is an integer greater than or equal to 1, and F creates a single static representation S(m,n) of the analog image. It is a halftoning function suitable for S (m, n) = F (m, n, {I (a, b), all a, all b}) image display method.